agentleFS
NatchpaiNatchpai/Hardware-Copilot/.claude/skills/hardware-checklist/SKILL.md

hardware-checklist

Skill0 starsChanged 17 days agoMITView on GitHub

Create, update, or audit an execution checklist for a hardware or PCB project. Turns specification requirements into gated tasks with owner, status, and required evidence. Use when the user asks for a hardware checklist, work plan, gate list, bring-up plan, or wants to check that every requirement has a task.

The file

Show its rules and text

We split this skill into the rules it gives an agent

23 rules. About 0 are common, 18 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈18

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use assets/execution_checklist_template.md when creating a new checklist.

    line 15
  2. Distinctive:

    Do not modify the template during normal work.

    line 17
  3. Distinctive:

    Create: Read the specification first.

    line 21
  4. Distinctive:

    Update: Read the existing checklist first.

    line 22
  5. Distinctive:

    Audit: Read-only. Compare the checklist with the specification and report:

    line 23
  6. Very specific:

    Task states: Use only: OPEN, IN_PROGRESS, BLOCKED BY <ID>, DEFERRED, DONE, N/A.

    line 59
  7. Distinctive:

    Owner: Every row has an owner role (Human owner, Hardware, Software, Fabricator, and so on).

    line 62
  8. Distinctive:

    Open items: Keep one short register that points to the specification's open-item IDs.

    line 65
  9. Very specific:

    Unknowns: Use TBD. Do not guess owners, dates, or evidence.

    line 66
  10. Distinctive:

    Use tables. Remove a column only if it is empty in every row.

    line 73
  11. Distinctive:

    Do not repeat the same fact in the checkpoint, the gate rows, and the register.

    line 74
  12. Distinctive:

    Prefer a few specific tasks over many vague ones.

    line 75
  13. Distinctive:

    Write simple, direct English, like one engineer talking to another.

    line 79
  14. Distinctive:

    Avoid fancy words, marketing phrases, and AI-style filler.

    line 80
  15. Distinctive:

    Keep standard engineering terms and exact IDs.

    line 81
  16. Very specific:

    Do not use the em dash (—).

    line 82
  17. Very specific:

    Use TBD for open values. Do not mix it with other spellings.

    line 83
  18. Distinctive:

    Do not open, parse, assess, or generate PDF content.

    line 88
  19. Distinctive:

    Keep KiCad design and library files read-only.

    line 89
  20. Distinctive:

    Do not run ERC or DRC, claim their results, or treat a pass as proof of correctness.

    line 90
  21. Distinctive:

    Do not mark a task done, approve a gate, or declare a phase for the human.

    line 91
  22. Distinctive:

    Do not stage, commit, or push.

    line 92
  23. Distinctive:

    Do not create logs, manifests, or other files that were not requested.

    line 93

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Hardware Checklist

Track the work and evidence needed to close each gate of a hardware or PCB project. The checklist is a tracker. It is not an approval record, a verification record, or a phase declaration.

Human engineers keep authority over decisions, approvals, and gate closure.

Template

  • Use assets/execution_checklist_template.md when creating a new checklist.
  • Adapt gates, columns, and rows to the project. Remove gates and rows that do not apply.
  • Do not modify the template during normal work.

Modes

  • Create: Read the specification first. Draft one or more tasks for each requirement or open item that needs work or evidence. Show the draft before writing the file, unless the user asked for the file.
  • Update: Read the existing checklist first. Change only the rows the user names. Keep IDs, order, and unrelated rows. Never reuse a removed ID.
  • Audit: Read-only. Compare the checklist with the specification and report:
    • requirement IDs with no task
    • tasks with no requirement or reason
    • DONE rows without evidence
    • open items without an owner or gate
    • rows that conflict with the specification
    • status cells that do not use the defined task states
    • open-item IDs that do not match the specification

Audit Output

Report ID coverage and topic coverage separately. ID coverage is a script or exact-match count and is VERIFIED. Topic coverage (a requirement is covered by a task that has no ID link) is a judgment and is ASSUMED. Do not mix the two.

  1. A counts table. Use a script for the counts when the files are large.

    CheckResult
    Spec requirement IDs linked from the checklistN of M
    Spec decision IDs linkedN of M
    Spec open-item IDs linkedN of M
    Task rows with an ownerN of M
    DONE rows with a named evidence recordN of M
  2. One line per finding, in this format:

    SEVERITY | ref | label | statement | source

    • Severity: MAJOR (requirement with no task, or DONE with no evidence), MINOR (missing detail or mismatch), INFO (wording or format). CRITICAL is not used here.
    • Label: VERIFIED (exact match or script count), ASSUMED (judgment), or NOT_VERIFIED (not checked).
    • Source: the script, or the spec and checklist sections read.
  3. A short list of proposed fixes. Do not apply them. The user chooses which to apply.

Rules

  1. Source of truth. Requirements and decisions come from the specification. Do not copy their values into the checklist. Reference the ID (for example PWR-001).
  2. Task states. Use only: OPEN, IN_PROGRESS, BLOCKED BY <ID>, DEFERRED, DONE, N/A. Define them once in a conventions block. Do not put evidence labels in the status column.
  3. Done needs evidence. Mark DONE only when the user states the task is complete and the evidence column names or links the record. A checkbox, file presence, or an ERC/DRC pass is not evidence by itself.
  4. Gates. Group tasks into gates. Each gate has an exit criterion. Default gates are P0 to P4: requirements, allocation, schematic, PCB and release package, bring-up. Gates are work stages. They are not the project phase (DRAFT, PRELIMINARY, CHECK, RELEASE). Never infer or change the project phase.
  5. Owner. Every row has an owner role (Human owner, Hardware, Software, Fabricator, and so on). Approval and gate-closure rows are owned by the human.
  6. Evidence required. Say what record closes the row: a table, a log, a review, a measurement. Be specific. Do not invent results.
  7. Deferred work. Mark non-hardware or later work DEFERRED and say who owns it. Say if it does not block the gate.
  8. Open items. Keep one short register that points to the specification's open-item IDs. Do not copy their descriptions.
  9. Unknowns. Use TBD. Do not guess owners, dates, or evidence.

Keep it short

  • Define status values once. Do not repeat them per row.
  • One task per row. Keep the task to one sentence.
  • Table format: pad the first column of every table to the width of its longest cell (header included), so the source lines up. Match the dashes in the separator row to that width. Leave the other columns unpadded.
  • Use tables. Remove a column only if it is empty in every row.
  • Do not repeat the same fact in the checkpoint, the gate rows, and the register. State it once.
  • Prefer a few specific tasks over many vague ones.

Writing style

  • Write simple, direct English, like one engineer talking to another. Use everyday words, short sentences, and active voice.
  • Avoid fancy words, marketing phrases, and AI-style filler.
  • Keep standard engineering terms and exact IDs.
  • Do not use the em dash (—). Use a period, comma, colon, or parentheses.
  • Use TBD for open values. Do not mix it with other spellings.
  • Priority: technical accuracy, then clarity, then brevity, then style.

Boundaries

  • Do not open, parse, assess, or generate PDF content.
  • Keep KiCad design and library files read-only.
  • Do not run ERC or DRC, claim their results, or treat a pass as proof of correctness.
  • Do not mark a task done, approve a gate, or declare a phase for the human.
  • Do not stage, commit, or push. Git actions need separate approval.
  • Do not create logs, manifests, or other files that were not requested.

Output

Report the file written, the rows added or changed, and any gap, such as a requirement with no task. Preserve existing project file names.

Suggested name: HARDWARE_CHECKLIST.md.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.